[Info-vax] Not getting a dump file when VMS crashes

abrsvc dansabrservices at yahoo.com
Tue Apr 3 10:52:15 EDT 2012


On Tuesday, April 3, 2012 10:18:57 AM UTC-4, Rich Jordan wrote:
> On Apr 2, 7:36 pm, David Froble <da... at tsoft-inc.com> wrote:
> > JF Mezei wrote:
> > > David Froble wrote:
> >
> > >> $ ana/crash
> > >> _Dump File: sys$system:sysdump.dmp
> >
> > > Don't you have to specify a dump file specification in SYSGEN  and a
> > > flag to cause VMS to dump to it ?
> >
> > Unless I've forgotten everything I've ever known (might not be much), the contents of
> > memory are written by default to SYS$SYSTEM:SYSDUMP.DMP.  I'm not trying to write the dump
> > to the paging file, or DOSD (off system disk) or anything like that.  My SYSDUMP.DMP is
> > the size of physical memory.
> 
> Also going from memory since its years since I tuned a VAX, but should
> not SYSDUMP.DMP be larger than actual memory by some increment when
> you're doing a full dump?  I don't think it was a big increment but it
> was there.
> 
> Ditto on the checking the fragmentation of the dump file; contiguous
> is best; we worked on (way back) a system that had run for two years
> straight and had its pagefile extended in chunks and a new CBT
> dumpfile made that ended up being in 20+ fragments that could not get
> a valid full dump (but did succeed with a selective dump).
> 
> Perhaps worth trying DUMPSTYLE=1 (or 3) and see if a selective dump
> succeeds and gives the info you need.

The dump file needs top be a little larger than mamory to allow for the errlog buffers.  I seem to recall the difference being around 100 blocks or so.

A selective dump should succeed as the most important info is saved first.  The dump file might be incomplete, but there should be enough info to get an idea of hte crash cause.

Dan



More information about the Info-vax mailing list